Fault Detection and Isolation for a Supermarket Refrigeration System - Part One: Kalman-Filter-Based Methods

Fault Detection and Isolation (FDI) using the Kalman Filter (KF) technique for a supermarket refrigeration system is explored. Four types of sensor fault scenarios, namely drift, offset, freeze and hard-over, are considered for two temperature sensors, and one type of parametric fault scenario, namely freeze-over/dirty built-up, is considered for one heat transfer coefficient between the inside air and the evaporator surface. For fault detection purpose, the fault residual is generated through a KF and then evaluated through CUSUM method. All fault scenarios can be detected clearly. For fault isolation purpose, a bank of KFs arranged by splitting measurements is constructed for sensor fault isolation, while the Multi-Model Adaptive Estimation (MMAE) method is employed to handle parametric fault isolation. All these approaches are extended and checked by using Extended KF technique afterwards. The test results show that the EKF-based FDI method generally performances better and faster than the KF-based method does. However, both methods can not handle the isolation between sensor faults and parametric fault.